Thursday, October 13, 2005

Knowledge freed from its husks

[In this entry and in all entries in this blog, selecting some text and releasing the mouse button will perform a Google search on that text.]

If you enter
capital Bhutan
into the Google search box, you will get at the top of the page

Bhutan — Capital: Thimphu
According to http://aardappel.co.uk/WebContent/Demographica/geos/bt.html - More sources »

You get the same thing if you enter: Bhutan capital.

My first question is: what sort of parsing does Google do? It certainly does some parsing of queries. If you enter an address, it recognizes it as an address. It apparently also recognizes UPS numbers, stock symbols, etc. For example, if you enter IBM, the first entry will be a stock quote and chart.

The point is not how clever Google is in parsing, or knowing where to go for stock quotes. It's the capital of Bhutan that I think is really significant.

I mentioned the talk recently by Pierre de Vries in which he talked about "digital technology sloughing off the husk of 'stuff' in which information has always been wrapped." Is this an initial example? Does Google recognize the question about Bhutan and know that the answer is somewhere on one of the pages it has scanned--but not necessarily which one? (With stock quotes, it knows in advance where to go.) If the page where it normally looks for "Capital Bhutan" is no longer available, will it look elsewhere? I tried the same search but restricted it not to use the original site. I didn't get an answer. But I'm not sure that's a good test.

If you look for
France president
the answer is retrieved from a different web page. The
Zimbabwe president
is retrieved from yet another site.

I think of this as a first sign of a new form of intelligence. Google (let's assume) can parse the query to mean "president of France." It can do enough parsing of the web sites it scans to know where that information is stored, and it can rank the reliability of those sites well enough to pick one as the place where the answer is given. In other words, the web is Google's database of knowledge. Google is not dependent on the precise structuring of that database. If that's the way it works, I find it pretty impressive. The knowlege has been freed from this husks that contain it. If one husk disappears, the knowledge will still be there in some other husk.

Whenever I'm talking with someone and a question comes up, my thought is "Ask Google; Google knows." What I normally mean is that with the appropriate queries, I can find out. But these examples are turning "Google knows" into something closer to the truth.

2 comments:

Anonymous said...

It's a neat trick, but that auto Google search when selecting text can get really annoying. Especially if someone just wants to copy a sentence or paragraph of your text.

Blue said...

I agree that it's annoying. It annoys me sometimes. A better engineered system is needed. Perhaps select with right mouse button or something like that. I don't have the time to work on it. The point of doing the auto-search was to illustrate the kinds of automatic linking one could put into a page. But as you say, every feature has its downside unless especially well engineered.